The approaches described in this section could be pursued, but are not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated herein, the approaches described in this section are not prior art to the claims in this application and are not admitted to be prior art by inclusion in this section.
Typically, networks today process only data packets and network data instead of application data. In a business-to-business environment, applications executing on computers commonly communicate with other applications that execute on other computers. For example, an application “A” executing on a computer “X” might send, to an application “B” executing on a computer “Y,” a message that indicates the substance of a purchase order.
Computer “X” might be remote from computer “Y.” In order for computer “X” to send the message to computer “Y,” computer “X” might send the message through a computer network such as a local area network (LAN), a wide-area network (WAN), or an inter-network such as the Internet. In order to transmit the message through such a network, computer “X” might use a suite of communication protocols. For example, computer “X” might use a network layer protocol such as Internet Protocol (IP) in conjunction with a transport layer protocol such as Transport Control Protocol (TCP) to transmit the message.
Assuming that the message is transmitted using TCP, the message is encapsulated into one or more data packets; separate portions of the same message may be sent in separate packets. Continuing the above example, computer “X” sends the data packets through the network toward computer “Y.” One or more network elements intermediate to computer “X” and computer “Y” may receive the packets, determine a next “hop” for the packets, and send the packets towards computer “Y.”
For example, a router “U” might receive the packets from computer “X” and determine, based on the packets being destined for computer “Y,” that the packets should be forwarded to another router “V” (the next “hop” on the route). Router “V” might receive the packets from router “U” and send the packets on to computer “Y.” At computer “Y,” the contents of the packets may be extracted and reassembled to form the original message, which may be provided to application “B.” Applications “A” and “B” may remain oblivious to the fact that the packets were routed through routers “U” and “V.” Indeed, separate packets may take different routes through the network.
A message may be transmitted using any of several application layer protocols in conjunction with the network layer and transport layer protocols discussed above. For example, application “A” may specify that computer “X” is to send a message using Hypertext Transfer Protocol (HTTP). Accordingly, computer “X” may add HTTP-specific headers to the front of the message before encapsulating the message into TCP packets as described above. If application “B” is configured to receive messages according to HTTP, then computer “Y” may use the HTTP-specific headers to handle the message. To construct an application-level message, the network device has to terminate the protocol and assemble the packets.
In addition to all of the above, a message may be structured according to any of several message formats. A message format generally indicates the structure of a message. For example, if a purchase order comprises an address and a delivery date, the address and delivery date may be distinguished from each other within the message using message format-specific mechanisms. For example, application “A” may indicate the structure of a purchase order using Extensible Markup Language (XML). Using XML as the message format, the address might be enclosed within “<address>” and “</address>” tags, and the delivery date might be enclosed within “<delivery-date>” and “</delivery-date>” tags. If application “B” is configured to interpret messages in XML, then application “B” may use the tags in order to determine which part of the message contains the address and which part of the message contains the delivery date.
Often, though, different applications are designed to use different application layer protocols to send and receive messages. For example, application “A” might be designed to send messages using only HTTP, but application “B” might be designed to receive messages using only File Transfer Protocol (FTP), another application layer protocol. Furthermore, different applications may be designed to use different message formats to format and interpret messages. For example, application “A” might be designed to format messages using only XML, but application “B” might be designed to interpret messages using only Electronic Data Interchange (EDI).
Usually, it is not practical or even possible to design or update an application so that the application can converse with other applications using all possible message formats and application layer protocols. Some message formats and application layer protocols may be proprietary and not publicly disclosed. Some message formats and application layer protocols may be relatively new and obscure. Some message formats and application layer protocols may be so old as to be considered generally obsolete.
In conventional practice, routers, switches, and other intermediary network elements route or switch individual frames, datagrams, and packets without any knowledge, awareness, or processing of the higher-order application layer messages embodied in flows of packets. In order to reduce the amount of application modification required to allow an application to converse with other applications that might use different message formats and/or application layer protocols, intermediary network elements separate from such applications may be designed to receive messages explicitly sent to those network elements, “translate” the messages, and then send the messages. This translation may be achieved by looking for a specified bit pattern beginning at a specified bit location in a packet, and then altering bits at the specified bit location if the specified bit pattern is found. For example, a network appliance “J” might be designed to receive messages that have been sent using HTTP and send those messages using FTP instead. For another example, a network appliance “K” might be designed to receive messages that are in XML format and translate those messages into EDI fornat. Thus, if application “A” sends messages in XML using HTTP, and application “B” receives messages in EDI using FTP, then application “A” can be configured so that messages that application “A” normally would address to application “B” are addressed to network appliance “J” instead. The network administrator can configure network appliance “J” to send messages to network appliance “K,” and the network administrator can configure network appliance “K” to send messages to application “B.”
Unfortunately, this approach requires a lot of effort from the network administrator. As the number of possible different application layer protocols and message formats used by communicating applications increases, the number of network appliances and paths between those network appliances rises dramatically. For each pair of sending and receiving applications, a network administrator following this approach must configure the applications and network appliances involved to ensure that the messages will follow the correct path through the relevant network appliances. Thus, if each of applications “A,” “B,” “C,” “D,” and “E” needed to communicate with each other, the network administrator following this approach might need to configure 25 different “paths” of one or more network appliances each. As applications are added, removed, and modified, the network administrator may need to add and/or remove certain network appliances from certain paths between application pairs. When many applications are involved, the burden can be more than most network administrators can bear.
Additionally, if multiple paths are configured to contain the same network appliance, then the network appliance may become a bottleneck that degrades network performance.
Thus, this “pair-wise path configuration” approach is impractical when applied to systems in which large numbers of diverse applications communicate. A more practical technique for allowing a multitude of diverse applications to communicate is needed.
Furthermore, existing intermediary network elements are limited in the kinds of processing that those network elements can perform on packets, and in the kinds of actions that those network elements can perform relative to packets. Typically, an existing intermediary network element performs only a single specialized operation relative only to data packets that possess matching parameters in those data packets' headers. Both the parameters and the operation are typically fixed and cannot be customized by an end user. A technique for allowing a greater number and variety of customizable operations to be performed relative to data flows is needed.
Additionally, processing the contents of data packets usually requires a significant amount of time and a significant portion of a network element's limited processing resources. In situations where network elements need to process the contents of data packets before sending those data packets to the data packets' destination, the network elements often cannot send the processed data packets at the same rate as the data packets would be sent if the contents of the data packets were not so processed. If a network element is forced to process the contents of all data packets that the network element receives, then packet latency may increase and packet throughput may decrease. Consequently, the overall performance of the network is generally degraded.
Under some circumstances, only the contents of some data packets need to be processed. Unfortunately, the determination of whether a particular data packet's contents need to be processed sometimes requires a thorough inspection of the contents themselves, and the thorough inspection requires nearly as significant an investment in time and processing power as the processing itself. A more efficient technique for reducing the fraction of data packets whose contents are thoroughly inspected is needed.